search

caoyunkang / Segment-Any-Anomaly

Official implementation of "Segment Any Anomaly without Training via Hybrid Prompt Regularization (SAA+)".
704 stars 76 forks source link

Domain Knowledge #14

Closed fujistoo closed 10 months ago

fujistoo commented 1 year ago

Hey! You've mentioned quite a few times "domain expert knowledge" in the paper, but you never mentioned anything about where exactly it came from, or how you filtered down to these prompts. Would you mind shedding some light onto this part?

caoyunkang commented 1 year ago

Hi! Typically, experts could easily introduce some descriptions of the potential anomalies, including anomaly types, anomaly properties, etc. In our implementation, we play the role of experts ourselves. In specific, we developed those prompts according to the given anomaly types and given images. We suggest that some pre-trained caption models could be also used to get more in-context descriptions rather than human descriptions.